
###################################################
# Replication of "BERT, RoBERTa or DeBERTa?
# by Joan C. Timoneda and Sebastian Vallejo Vera
###################################################

1. For application 1 (civility): 

For BERT, RoBERTa and DeBERTa results, run the file: 
	./Application 1 - Civility/Analysis/results_civility.py

For SVM and Bi-LSTM baseline results, run the files:
	./Application 1 - Civility/Baseline code/SVM_civility.py
	./Application 1 - Civility/Baseline code/BiLSTM_civility.py


2. For application 2 (speeches):

For mBERT, XLM-RoBERTa and mDeBERTa results, run the file: 
	./Application 2 - Speeches/Analysis/results_speeches.py

For SVM and Bi-LSTM baseline results, run the files:
	./Application 2 - Speeches/Baseline code/SVM_speeches.py
	./Application 2 - Speeches/Baseline code/BiLSTM_speeches.py


3. For application 3 (Fake News):

For mBERT, XLM-RoBERTa and mDeBERTa results, run the file: 
	./Application 3 - Fake news/Analysis/results_fakenews.py

For SVM and Bi-LSTM baseline results, run the files:
	./Application 3 - Fake news/Baseline code/SVM_speeches.py
	./Application 3 - Fake news/Baseline code/BiLSTM_speeches.py

############################################################################

Notes:

- Make sure you place the data and the
'glove_vectors_w2v.txt' files in the appropriate path. 
- Replication code for the supporting material is in the "Supporting 
material" folder. Run the file 'results_benchmark.py' in "Benchmarking".






